We provide tools to perform low-layer attacks such as reactive and constant jamming using commodity devices. Reactive jamming allows you to block specific Wi-Fi packets. For example, all beacons and probe responses of a specific Access Point (AP) can be jammed. It has been tested with the following devices:
This work was the result of the paper [Advanced Wi-Fi Attacks Using Commodity Hardware](https://lirias.kuleuven.be/bitstream/123456789/473761/1/acsac2014.pdf) presented at ACSAC 2014. *If you use these tools in your research, please reference this paper.* Most code is open source, and contributions are welcome. The code of the constant jammer can be requested but is not available publicly. Don't worry, we won't bite.
April 2016: we now support Linux kernels 3.0 up to and including 4.4! See the [modwifi-4.4-1.tar.gz](releases/modwifi-4.4-1.tar.gz) release! This has been tested on Arch Linux and Ubuntu 15.10.
You can [download a VMWare image](http://people.cs.kuleuven.be/~mathy.vanhoef/modwifi/Xubuntu-Modwifi.7z) that has the drivers, firmware, and user-land tools preinstalled. Just boot it, plug-in the USB dongle, and start experimenting! **The password of the account modwifi is modwifi**. Once booted, you can execute (the public) attacks below.
## Basic Usage
This section describes the attacks that can be executed. We assumed you already downloaded the VMWare image or manually installed the drivers and firmware (see the section "Installation" to install drivers on your existing machine).
*Before doing any attacks it is recommended to disable WiFi.* In particular I mean disabling WiFi in your network manager. Most graphical network managers have an option somewhere named "Enable Wi-Fi". Make sure it's not selected. If you can't find it, perhaps you can disable in the terminal with `sudo nmcli nm wifi off`. Once you have disabled WiFi your OS won't interfere with our attacks.
*If RF-kill is enabled* we'll have to turn it off. Some distributions set RF-kill on after disabling WiFi. But we still want to actually use our WiFi devices. So execute:
Our current implementation of our reactive jammer allows you to block an Access Point. More precisely, all beacons and probe responses will be jammed. Execute it using:
**The first three commands need to be executed only once** after plugging in your dongle. To get the interface name of the wireless card you can execute `iwconfig`. In this case our targeted AP was on channel 11, but remember that your targeted AP may be on a different channel.
You can stop the reactive jammer using CTRL+C. It may take a few seconds before it actually stops. By modifying the firmware you can reactive jam any kind of packets you like. For example, you could jam all packets of a specific client. Note that only medium to large packets can be reliably jammed (see our paper).
You can verify that this works by monitoring the channel with a second device. Make sure that this device also reports corrupted frames using:
```bash
sudo iw wlan1 set monitor fcsfail
```
This will instruct the driver to also pass corrupted frames to the userland (when in monitor mode). Be warned though, not all drivers properly support this flag. Some will always show corrupted frames. Others will never show corrupted frames. Our drivers and firmware handle this flag correctly!
Want to disable carrier sense in order to perform an experiment? Then execute this:
```bash
modwifi@ubuntu:~$ sudo su
root@ubuntu:~$ mount -t debugfs none /sys/kernel/debug
root@ubuntu:~$ cd /sys/kernel/debug/ieee80211/phy*/ath9k_htc/registers/
root@ubuntu:~$ echo 1 > force_channel_idle
root@ubuntu:~$ echo 1 > ignore_virt_cs
```
Writing 1 to `force_channel_idle` disables physical carrier sense (channel is busy). Writing 1 to `ignore_virt_cs` disables virtual carrier sense (RTS/CTS). Random backoff parameters can also be changed.
This performs constant jamming on channel 6. Because channels overlap, nearby channels will also be jammed. Remember that the constant jamming implementation is not public, but can be requested privately.
The specific scripts we used to easily configure a device to act unfairly are not public. The reason behind this is that it's hard to defend against these kind of attacks. However, some parameters can still be accessed as `debugfs` entries in `/sys/kernel/debug/ieee80211/phy*/ath9k_htc/registers/`.
This is an advanced attack and not for the fainthearted. It clones an existing Access Point on a different channel. This allows us to reliably manipulate encrypted traffic. We used this to break TKIP. See [our paper]() for details. An example on how we used it to verify that our awesome-sauce attacks work:
- You can only change the channel of a monitor device when no other (virtual) interface is active. So if you have a `monX` interface, you need to bring down (`ifconfig wlanX down`) all other interfaces (which use that device) first.
- In general you want to kill other processes that are trying to use/configure your WiFi device. Tools like [`airmon-zc`](http://svn.aircrack-ng.org/tags/1.2-beta3/manpages/airmon-zc.8) can help detect which processes might be interfering. Note that `airmon-zc` is the successor of the older `airmon-ng` tool.
You can also install the latest drivers and firmware on your own machine. The quickest method is to grab [one of our release packages](https://github.com/vanhoefm/modwifi/raw/master/releases/). Only your wireless stack and drivers will be replaced, all other drivers will remain the same (if you use other wifi devices as well, compile them too). Normal usage of WiFi still works perfectly when these drivers are installed (I use these drivers myself :).
Note that this only compiles and installs the ath9k drivers. If you want to use modwifi, and at the same time control other wireless networks cards on the kernel, modify and use the appropriate `defconfig-*` file (e.g. include the appropriate flags in `defconfig-ath9k-debug` so the drivers you need are also compiled).
If you want to compile the firmware as well, clone the [ath9k-htc repository](https://github.com/vanhoefm/modwifi-ath9k-htc), and follow the instructions there. If you want to modify the driver, you can modify the downloaded code in `modwifi-YYYYMMDD.tar.gz`. You can put that code in your own repository to keep track of changes, and send us patches based on this. Alternatively, the more correct but also significantly more tedious method, would be to clone the research branch of [our forked Linux kernel](https://github.com/vanhoefm/modwifi-linux). The driver can be extracted from the kernel code using the [backports](https://backports.wiki.kernel.org/index.php/Main_Page) project. You can then install the drivers only (so without modifying your own kernel).
Our drivers and firmware can be run on a Raspberry Pi. We tested this using raspbian. In order to get it working first download and update some dependencies:
As you can see, we tested this on the 3.12-1-rpi kernel. You can use another kernel if you want, just be sure to download the kernel headers. To enable the 3.12-1-rpi kernel we just downloaded edit `/boot/config.txt` and append:
kernel=vmlinuz-3.12-1-rpi
initramfs initrd.img-3.12-1-rpi followkernel
And to assure our raspberry pi will recognize the device when we plug it in, execute:
Everything is now ready to install our drivers and firmware. Just **follow the instructions under section "Installation"**. Compilation of the drivers can take a while. Finally we have to prevent raspbian from automatically trying to enable and manage WiFi (this interferes with our attacks). First edit `/etc/network/interfaces` and comment out the following two lines:
Now edit `/etc/default/ifplugd` and change the `INTERFACES` and `HOTPLUG_INTERFACES` to:
INTERFACES="eth0"
HOTPLUG_INTERFACES="eth0"
ARGS="-q -f -u0 -d10 -w -I"
SUSPEND_ACTION="stop"
This will prevent raspbian from automatically enabling and managing the wireless interface (so we can first put the device in monitor mode and only then enable it). You can now compile the tools and execute the attacks!
To compile the Linux and ath9k-htc firmware, read the documentation of these projects. To backport the modified drivers using the backports project, also see the official documentation of that project. Finally, our tools can be compiled using a simple `make all`. Apart from the `tools` repositories, all work and modifications are performed on the `research` branch. When a new Linux kernel (or firmware) is released, we can easily merge with it. As a result **our code is relatively easy to keep up-to-date**.
For those who also want to start hacking away at the driver and firmware, I recommend first reviewing our patches. This allows you to study what our changes do, and inspect the firmware code at small chunks one at a time. That way it's easier to learn step by step. Maybe you will even find bugs or can make improvements (let us know). Also, in the `ath9k-htc` repository, there is a directory called `docs`. While still terse to read, these documents should be an excellent guide while reading and understanding the code.
We rely on the [driver backports project](https://backports.wiki.kernel.org/index.php/Main_Page) to ship our modified drivers (as installable modules) to older kernels. This provides two advantages: (1) the drivers can be installed as modules, meaning user don't have to (re)compile a linux kernel; and (2) these drivers (i.e., modules) are compatible with many recent kernel versions. Officially the [backports project tracks the linux-next tree](https://backports.wiki.kernel.org/index.php/Documentation/backports/hacking#Git_trees_you_will_need). This means it extracts, and backports, recents drivers (as loadable modules) from the linux-next tree. However, we base our code on [Linus' tree](http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git), because [basing code on linux-next is not really possible](https://lwn.net/Articles/289013/).
To target a new kernel there are two cases:
1. There is a `linux-x.y.z` branch of the kernel you want to target. Checkout this branch, use it to extract modified drivers. Should be relatively simple.
2. We need to create our own `linux-x.y.z` branch. I call my own batches of this type `mathy-x.y.z`. It is best to first get backports working against a linux-next tag that was tested by the backports project itself. Then I found that, with some minor patches, backports will also work against Linus' tree of a specific version tag that is close to the tested linux-next snapshot. So find the latest linux-next snapshot that is compatible with the backports project. [How to get linux-next](https://www.kernel.org/doc/man-pages/linux-next.html). You may need to use [linux-next-history](http://git.kernel.org/cgit/linux/kernel/git/next/linux-next-history.git) for this, which [contains all linux-next tags](http://lkml.iu.edu/hypermail/linux/kernel/1108.0/00555.html) ever created. It may be that backports cannot cleanly extract the driver code. If so, crease a new research branch in the backports repository, and modify the patches so everything cleanly applies and compiles.
**Use python2 when using backports**. Some scripts may fail if python3 is the default.
Note that branches named `mathy-x.y.z` are custom branches, I personally use them to create backports for my _currently running_ linux kernel. For example, `mathy-4.7.y` can take code from linux-next, and will compile on Linux 4.7.4. However, it may not compile on older kernels! While working on backports, you may find it useful to use `rediff` from the `patchutils` package to manually change patch files.
